BNA score for Europe

An attempt to reproduce the People for Bikes BNA score

The codes below will allow the computation of the BNA score for any city in Europe. The coding languages are a combination of R and SQL, with a few lines passed onto the Command Prompt.

Workflow step by step

1. Establish a connection with the PostgreSQL server and establish basic parameters of analysis.

Within this step, three major things are performed, assuming that the user has created a database on its PostgreSQL. To test conneciton an empty table on the database called “test” was created on the public schema.

- Create connection and test it

library(RPostgreSQL)

Loading required package: DBI

# LOAD POSTGRESQL DRIVER
driver <- dbDriver("PostgreSQL")
# CREATE CONNECTION TO THE POSTGRESQL DATABASE
# THE CONNECTION VARIABLE WILL BE USED FOR ALL FURTHER OPERATIONS
connection <- dbConnect(
  driver, 
  dbname = db_name,
  host = local_host, 
  port = port_num,
  user = user_name, 
  password = rstudioapi::askForPassword("Database password")
)
ifelse(
  !dbExistsTable(connection, "test"),
  "The connection to the database was not possible.",
  "The connection to the database was successful!"
)

[1] "The connection to the database was successful!"

# DISCONNECT: Important when modifying the database on pgadmin4
# dbDisconnect(connection)

- Set up database

NOTICE:  extension "hstore" already exists, skipping
NOTICE:  extension "postgis" already exists, skipping
NOTICE:  extension "pgrouting" already exists, skipping
NOTICE:  schema "destinations" already exists, skipping
NOTICE:  schema "generated" already exists, skipping
NOTICE:  schema "received" already exists, skipping

- Establish Study Area and other important variables

Not only the name of the study area should be established, but also the number of subdivisions for the grid, the coordinate reference system to work with, and the biking distance that will be assumed for the connectivity analysis.

sa_name = "Cambridge"
subdivisions = 4
sa_crs = 3857
biking_distance = 3000 ## in meters
paste("You are running the BNA score for", sa_name, "within a biking distance of", biking_distance/1000, "km.")

[1] "You are running the BNA score for Cambridge within a biking distance of 3 km."

2. Obtain study area boundary with osmdata.

# FUNCTION TO EXTRACT STUDY AREA BOUNDARY WITH OSM AND WRITE IT TO THE DATABASE
sa_bb <- function (study_area, crs, conn){
  # GET DATA FROM OSM
  library(osmdata)
  library(sf)
  study_area_bb <- study_area %>% 
    getbb(format_out = "sf_polygon") %>%
    st_transform(crs = crs)
  
  # DELETE EXISTING BOUNDARY
  library(sqldf)
  sqldf(
    "DROP TABLE IF EXISTS received.sa_boundary",
    connection = conn
  )
  # UPLOAD BOUNDARY TO POSTGRESQL DATABASE
  library(RPostgreSQL)
  dbWriteTable(conn, c("received","sa_boundary"), study_area_bb)
  
  study_area_bb
}
boundary <- sa_bb(
  study_area = sa_name, 
  crs = sa_crs, 
  conn = connection
)

Data (c) OpenStreetMap contributors, ODbL 1.0. http://www.openstreetmap.org/copyright
Linking to GEOS 3.6.1, GDAL 2.2.3, proj.4 4.9.3
Loading required package: gsubfn
Loading required package: proto
Loading required package: RSQLite
sqldf will default to using PostgreSQL

library(tmap)
tmap_mode("view")

tmap mode set to interactive viewing

qtm(
  shp = boundary, 
  fill = NULL, 
  borders = "red", 
  basemaps = "OpenStreetMap"
)

3. Obtain the .osm file from Overpass API

# FUNCTION TO DOWNLOAD OSM DATA WITH THE OVERPASS API 
sa_download <- function(conn){
  # OBTAIN THE EXTENT OF THE STUDY AREA AS A BOUNDING BOX
  sa_extent <- dbGetQuery(conn,
        "SELECT
        ST_Extent((ST_Transform(geometry,4326)))
        FROM received.sa_boundary") 
  
  library(stringr)
  sa_coord <- toString(sa_extent) %>% 
    str_extract_all("\\-*\\d+\\.*\\d*") %>% 
    unlist() %>% 
    toString()
    
  # CONSTRUCT THE API LINE TO REQUEST THE DATA
  api <- paste(
    'https://overpass-api.de/api/map?bbox=',
    sa_coord,
    sep = ''
  )
  
  # CREATE A NEW TEMPORAL DIRECTORY TO DOWNLOAD THE INFO
  cd <- getwd()
  ifelse(
    !file.exists(file.path(cd,'temp')),
    dir.create(file.path(cd,'temp')), 
    "Directory already exists"
  )
  
  # ESTABLISH THE NAME OF THE FILE WHERE THE OVERPASS API WILL DOWNLOAD ITS DATA
  osm_file <- file.path(cd,'temp','overpass.osm')
  
  # REQUEST THE DATA FROM THE API
  library(utils)
  download.file(url = api, destfile = osm_file, extra = '-nv -O') 
  
  ifelse(
    file.exists(file.path(cd,'temp','overpass.osm')),
    "OMS data successfully downloaded!",
    "OSM data was not downloaded, please try again or download manually."
  )
  
}
# DOWNLOAD THE DATA FROM OSM WITH OVERPASS API
sa_download(conn = connection)

RS-DBI driver warning: (unrecognized PostgreSQL field type box2d (id:56298) in column 0)trying URL 'https://overpass-api.de/api/map?bbox=0.0686389, 52.1579417, 0.184552, 52.2372296'
Content type 'application/osm3s+xml' length unknown
downloaded 75.1 MB

[1] "OMS data successfully downloaded!"

4. Load data into the PostgreSQL database.

- Obtain configuration files

# CREATE A NEW TEMPORAL DIRECTORY TO DOWNLOAD THE INFO
cd <- getwd()
ifelse(
  !file.exists(file.path(cd,'temp')),
  dir.create(file.path(cd,'temp')), 
  "Directory already exists"
)

[1] "Directory already exists"

# ESTABLISH THE NAME OF THE FILES 
pfbstyle_file <- file.path(cd,'temp','pfb.style')
mapconfig_file <- file.path(cd,"temp","mapconfig.xml")
mapconfigbikes_file <- file.path(cd,"temp","mapconfig_for_bicycles.xml")
# CHECK IF THEY ARE DOWNLOADED
if(
  file.exists(file.path(pfbstyle_file)) & 
  file.exists(file.path(mapconfig_file)) &
  file.exists(file.path(mapconfigbikes_file))
){
  "Files are already downloaded!"
} else {
 # ESTABLISH THE URLS
  pfbstyle_url <- "https://raw.githubusercontent.com/azavea/pfb-network-connectivity/develop/src/analysis/import/pfb.style"
  
  mapconfig_url <- "https://raw.githubusercontent.com/pgRouting/osm2pgrouting/master/mapconfig.xml"
  
  mapconfigbikes_url <- 
  "https://raw.githubusercontent.com/pgRouting/osm2pgrouting/master/mapconfig_for_bicycles.xml"
  
  # REQUEST THE DATA
  library(utils)
  download.file(url = pfbstyle_url, destfile = pfbstyle_file)
  download.file(url = mapconfig_url, destfile = mapconfig_file)
  download.file(url = mapconfigbikes_url, destfile = mapconfigbikes_file) 
}

trying URL 'https://raw.githubusercontent.com/azavea/pfb-network-connectivity/develop/src/analysis/import/pfb.style'
Content type 'text/plain; charset=utf-8' length 11639 bytes (11 KB)
downloaded 11 KB

trying URL 'https://raw.githubusercontent.com/pgRouting/osm2pgrouting/master/mapconfig.xml'
Content type 'text/plain; charset=utf-8' length 1861 bytes
downloaded 1861 bytes

trying URL 'https://raw.githubusercontent.com/pgRouting/osm2pgrouting/master/mapconfig_for_bicycles.xml'
Content type 'text/plain; charset=utf-8' length 3013 bytes
downloaded 3013 bytes

- Load data with osm2pgsql and osm2pgrouting

NOTE: To run this command create a password file on %APPDATA%/postgresql/pgpass.conf with the format hostname:port:database:username:password

Replace variables between %:

system(
  command = "osm2pgsql -c -d %DBNAME% -U %USERNAME% -H %HOSTNAME% -W --create --prefix sa_full -S %CURRENTDIRECTORY/temp/pfb.style% %CURRENTDIRECTORY/temp/overpass.osm% --cache 600",
  show.output.on.console = TRUE
  )

system(
  command = "osm2pgrouting -f %CURRENTDIRECTORY/temp/overpass.osm% -h %HOSTNAME% --password %DBPASSWORD% -d %DBNAME% --username %USERNAME% --schema received --prefix sa_all_ --conf %CURRENTDIRECTORY/temp/mapconfig.xml% --clean",
  show.output.on.console = TRUE
)

system(
  command = "osm2pgrouting -f %CURRENTDIRECTORY/temp/overpass.osm% -h %HOSTNAME% --password %DBPASSWORD% -d %DBNAME% --username %USERNAME% --schema received --prefix sa_bike_ --conf %CURRENTDIRECTORY/temp/mapconfig_for_bicycles.xml% --clean",
  show.output.on.console = TRUE
)

5. Organize and prepare the database

On this step, several SQL queries are being run to organize the tables, clip them to the study area outline, merge it with the osm2pgsql data among other things. The scripts include mainly the code that PfB already uses, but with some modifications like changing feet to meters, mph to km/h, and others. Basically, 4 steps are taken:

- Organize tables

Which drops unused columns and projects data to the already established CRS. It also cleans the database for a new analysis to be run if the study area is changed for example.

- Clip data into boundary

Which clips the data into the boundary of the study area selected.

- Populate way table

Which does modifications to the following columns on the ways table:

> one way

> width

> functional class

> paths

> speed limit

> lanes

> park

> bike infrastructure

> class adjustments

- Populate intersection table

Which does modifications to the following columns on the intersection table.

> legs

> signalized

> stops

> rrfb

> island

An example of how the tables look like after this step:

SELECT * FROM received.sa_ways LIMIT 10;

SELECT * FROM received.sa_ways_int LIMIT 10;

The code for this step is not included as it is basically the same as the PfB code, and is actually quite long. However, it can be examined through the Rmd file for this R Notebook.

One additional value that I was considering on adding is slope, however I have not gone through with the complete implementation of the variable yet. This would also affect step 6.

6. Calculate stress

The same as step 5, this step considers mainly SQL queries already performed by PfB. It will mainly alter the columns meant to host the stress rank for segments and intersections. It considers different cases to do the classification. The scripts that can be further examined on the Rmd file are:

> motorway trunk

> higher order

> lower order

> living street

> track

> path

> one way reset

> motorway trunk intersection

> primary intersection

> secondary intersection

> tertiary intersection

> lower intersection

> link intersection

An example of how the tables look like after this step:

SELECT osm_id, name, ft_seg_stress, tf_seg_stress, ft_int_stress, tf_int_stress FROM received.sa_ways LIMIT 10;

7. Build network

On this step the network is built by creating two tables: vertices and links. As the last 2 steps, the code won’t be include but can be analyzed on the Rmd file.

8. Generate population grid

So, this is one of the main differences regarding the PfB approach and mine. Instead of using US census blocks I used a population grid of 1 km² for the entire European territory. Since its area is quite big, I created a subdivision code to split the data, considering partial populations for each new cell depending on the mother cell. To do this I followed two steps:

- Download data from EUROSTAT and load into DB

## Download data and load to PostgreSQL
if (!dbExistsTable(connection, c("received","geostat"))){
  
  # CREATE A NEW TEMPORAL DIRECTORY TO DOWNLOAD THE INFO
  cd <- getwd()
  ifelse(
    !file.exists(file.path(cd,'temp')),
    dir.create(file.path(cd,'temp')), 
    "Directory already exists"
  )
  
  # ESTABLISH THE NAME OF THE FILE WHERE THE GEOSTAT DATA WILL BE DOWNLOADED AND UNZIPPED
  geostat_file <- file.path(cd,'temp','geostat.zip')
  geostat_exdir <- file.path(cd,"temp","geostat")
  
  if (!file.exists(geostat_exdir)){
    # DEFINE THE URL FROM WHERE THE DATA COMES
  
    geostat_url <- 
      "https://ec.europa.eu/eurostat/cache/GISCO/geodatafiles/GEOSTAT-grid-POP-1K-2011-V2-0-1.zip"
    
    # DOWNLOAD THE FILE, UNZIP IT AND DELETE .ZIP
    
    library(utils)
    download.file(url = geostat_url, destfile = geostat_file)
    unzip(geostat_file, exdir = geostat_exdir)
    file.remove(geostat_file)
  }
  
  # CALL DATA INTO R AND REPROJECT
  library(sf)
  
  table_path <- file.path(
    geostat_exdir,
    "Version 2_0_1/GEOSTAT_grid_POP_1K_2011_V2_0_1.csv"
  )
  
  grid_path <- file.path(
    geostat_exdir,
    "Version 2_0_1/GEOSTATReferenceGrid/Grid_ETRS89_LAEA_1K-ref_GEOSTAT_POP_2011_V2_0_1.shp"
  )
  
  pop_table <- st_read(table_path)
  names(pop_table) <- pop_table %>% names() %>% tolower()
  
  pop_grid <- st_read(grid_path)
  pop_grid_t <- pop_grid %>% st_transform(crs = sa_crs)
  names(pop_grid_t) <- pop_grid_t %>% names() %>% tolower()
  
  # LOAD TO POSTGRESQL
  library(sqldf)
  sqldf(
    "
  DROP TABLE IF EXISTS received.pop_grid;
  DROP TABLE IF EXISTS received.pop_table;
    ",
  connection = connection
  )
  
  dbWriteTable(
    conn = connection,
    name = c("received","pop_grid"),
    value = pop_grid_t
  )
  
  dbWriteTable(
    conn = connection,
    name = c("received","pop_table"),
    value = pop_table
  )
  
  #### Join tables on data base and extract study area
  
  sqldf(
    "
-- Create join between .csv and .shp
DROP TABLE IF EXISTS received.geostat;
DROP INDEX IF EXISTS received.geostat_geom_idx;
CREATE TABLE received.geostat AS
 SELECT grid.grd_id, grid.geometry, tab.tot_p, tab.cntr_code, tab.year, tab.tot_p_con_dt
    FROM received.pop_grid grid, received.pop_table tab
    WHERE grid.grd_id = tab.grd_id;
CREATE INDEX geostat_geom_idx
  ON received.geostat
  USING gist
  (geometry);
  
DROP TABLE IF EXISTS received.pop_grid;
DROP TABLE IF EXISTS received.pop_table;
    ",
    connection = connection
  )
} else {
  "GEOSTAT data already loaded to database."
}

trying URL 'https://ec.europa.eu/eurostat/cache/GISCO/geodatafiles/GEOSTAT-grid-POP-1K-2011-V2-0-1.zip'
Content type 'application/zip' length 60101856 bytes (57.3 MB)
downloaded 57.3 MB

Reading layer `GEOSTAT_grid_POP_1K_2011_V2_0_1' from data source `E:\GeoTech\Thesis\Thesis_R_Project\temp\geostat\Version 2_0_1\GEOSTAT_grid_POP_1K_2011_V2_0_1.csv' using driver `CSV'

no simple feature geometries present: returning a data.frame or tbl_df

Reading layer `Grid_ETRS89_LAEA_1K-ref_GEOSTAT_POP_2011_V2_0_1' from data source `E:\GeoTech\Thesis\Thesis_R_Project\temp\geostat\Version 2_0_1\GEOSTATReferenceGrid\Grid_ETRS89_LAEA_1K-ref_GEOSTAT_POP_2011_V2_0_1.shp' using driver `ESRI Shapefile'
Simple feature collection with 2106228 features and 1 field
geometry type:  POLYGON
dimension:      XY
bbox:           xmin: 944000 ymin: 942000 xmax: 6503000 ymax: 5414000
epsg (SRID):    NA
proj4string:    +proj=laea +lat_0=52 +lon_0=10 +x_0=4321000 +y_0=3210000 +ellps=GRS80 +units=m +no_defs

NOTICE:  table "pop_grid" does not exist, skipping
NOTICE:  table "pop_table" does not exist, skipping
NOTICE:  table "geostat" does not exist, skipping
NOTICE:  index "geostat_geom_idx" does not exist, skipping

Warning message:
In postgresqlExecStatement(conn, statement, ...) :
  RS-DBI driver warning: (unrecognized PostgreSQL field type geometry (id:56267) in column 5)

- Generate subdivision with a grid, adding partial population and unique ID

sqldf::sqldf(
  "
DROP TABLE IF EXISTS received.sa_geostat;
DROP INDEX IF EXISTS received.sa_geostat_geom_idx;
-- Extract the grids concerning only the study area
CREATE TABLE received.sa_geostat AS
    SELECT  DISTINCT geo.grd_id, 
            CAST(geo.tot_p AS INTEGER), 
            geo.cntr_code, 
            geo.geometry
    FROM received.geostat geo, received.sa_ways w
    WHERE ST_Intersects(geo.geometry, w.geom);
CREATE INDEX sa_geostat_geom_idx
  ON received.sa_geostat
  USING gist
  (geometry);  
  ",
  connection = connection 
)

NOTICE:  table "sa_geostat" does not exist, skipping
NOTICE:  index "sa_geostat_geom_idx" does not exist, skipping

## Establish a function to create grid with different number of subdivisions, defaults to 9
grid <- function(s = 9){
  ## Call it as an sf object and then transform it to CRS:3035 to create grid
  library(sf)
  library(dplyr, quietly = TRUE)
  
  sa_pop_1km2 <- st_read(
    dsn = connection,
    layer = c("received", "sa_geostat")
  ) %>% 
    st_transform(crs = 3035)
  
  ## Determine number of horizontal and vertical cells
  h <- as.integer(as.numeric(diff(st_bbox(sa_pop_1km2)[c(1, 3)]))/1000)
  v <- as.integer(as.numeric(diff(st_bbox(sa_pop_1km2)[c(2, 4)]))/1000)
  
  ## Make grid
  grid <- sa_pop_1km2 %>% 
    st_make_grid(n=c(h*sqrt(s),v*sqrt(s)), what = "polygons") %>%
    st_sf() %>% 
    mutate(id = 1:n()) %>% 
    st_intersection(sa_pop_1km2)
  
  ## Filter grid by area of intersection because there are small polygons created.
  
  grid$area <- grid %>% st_geometry() %>% st_area() %>% as.numeric()
  
  grid <- grid %>% filter(area > 1)
  
  grid$area <- NULL
  
  grid <- within(grid, cell_id <- paste(grd_id,"C",id, sep = ""))
  
  grid$id <- NULL
  
  grid$partial_p <- grid$tot_p/s
  
  grid %>% st_transform(crs = sa_crs)
}
sa_grid <- grid(s = subdivisions) # Always consider a squared number to make an even division

Attaching package: 㤼㸱dplyr㤼㸲

The following objects are masked from 㤼㸱package:stats㤼㸲:

    filter, lag

The following objects are masked from 㤼㸱package:base㤼㸲:

    intersect, setdiff, setequal, union

attribute variables are assumed to be spatially constant throughout all geometries

sqldf::sqldf(
  "DROP TABLE IF EXISTS generated.sa_pop_grid",
  connection = connection
)

NOTICE:  table "sa_pop_grid" does not exist, skipping

## Load data into data base
RPostgreSQL::dbWriteTable(
  conn = connection,
  name = c("generated","sa_pop_grid"),
  value = sa_grid
)

[1] TRUE

library(tmap)
tmap_mode("view")

tmap mode set to interactive viewing

qtm(
  shp = sa_grid, 
  fill = NULL, 
  borders = "red", 
  basemaps = "OpenStreetMap"
)

Plotting the grid on this step can allow the analyst decide on a better number of subdivisions, depending on the study area. I hope to automatize this on a later effort.

- Prepare the population grid table

This step is only generating new columns on my new Population Grid table. It follows the same logic as the PfB and therefore won’t be inlcuded on this document explicitely.

9. Reachable roads scripts

This is the core of the whole BNA analysis, where the actual network analysis is performed. This step might take some computation time. It is again the same as PfB, and can be reviewed with more detail on the Rmd file. It basically uses pgrouting, therefore the Dijkstra algorithm to compute the driving distance considering the configuration established on step 7. It does it for the two levels of traffic stress:

> High stress

> Low stress

10. Establish connected population grids and compute their accessibility

On this step basically four procedures take place:

- Connect population grids

Where a new table sa_connected_pop_grid is created to summarize the connected cells by establishing them as source and target, including if they are connected by the low or high stress network, and obtaining the minimum the costs between cells.

- Compute population access

The access computation on this step fills up the sa_pop_grid table created on step 8, according to the PfB methodology.

To compute access on this an the next step, a weighting procedure is used, as the methodology of PfB does, which can be accessed here.

A quick glance of the weights used, mainly for step 11:

Scoring Category	Measure
People = 15	Population = N/A
Opportunity = 20	Employment = 35
	K-12 Education = 35
	Technical/vocational school = 10
	Higher Education = 20
Core Services = 20	Doctor offices/clinics = 20
	Dentist offices = 10
	Hospitals = 20
	Pharmacies = 10
	Supermarkets = 25
	Social services = 15
Recreation = 15	Parks = 40
	Recreational trails = 35
	Community centers = 25
Retail = 15	Retail shopping = N/A
Transit = 15	Station/transit centers = N/A

It is important to note that this reproduction of the BNA for Europe does not include employment data, as until now, I have not located a source to provide this information as open data for the whole Europe. Therefore, the final results will show this category but with 0 or NA values.

- Extract common destinations

Which uses the osm polygons and points generated by osm2pgsql. The destinations included are:

> Colleges

> Community centers

> Dentists

> Doctors

> Hospitals

> Parks

> Pharmacies

> Retail

> Schools

> Social services

> Supermarkets

> Transit

> Universities

- Compute access to common destinations

Where the access to the destinations established before is computed. Access to recreational trails and bike paths is also included.

Once again, on this step I do not include the SQL codes, however they can be accessed through the Rmd file.

11. Compute overall access

During this step the overall access is computed for each population grid, meaning that we can already observe the BNA score spatial behavior within our study area.

12. Compute overall score for the whole study area

For this step a new table is generated in the database sa_score_inputs to store the preliminary results. The code can be accessed on the Rmd file.

Results

The overall results obtained include the final score for the whole city, as well as the score per destination category. The total population and stress network is also calculated. The results can be observed on the following table.

	Score/Value
Overall Score	74.32
Population	130217
Length of Low Stress Network (km)	1286.6
Length of High Stress Network (km)	118.3
People
Total People	79.65
Opportunity
Employment	0
K-12 Education	84.47
Technical/vocational school	81.24
Higher Education	77.46
Total Opportunity	53.18
Core Services
Doctor offices/clinics	81.12
Dentist offices	78.79
Hospitals	72.68
Pharmacies	77.83
Supermarkets	87.17
Social services	84.89
Total Core Services	80.95
Retail
Total Retail shopping	81.62
Recreation
Parks	86.26
Recreational trails	95.76
Community centers	73.44
Total Recreation	86.38
Transit
Total Transit	68.97

We can plot the results to have a quick view of the output, including the high and low stress network in an interactive way.

library(sf)
pop <- st_read(
  dsn = connection,
  layer = c("generated","sa_pop_grid")
)
ways <- st_read(
  dsn = connection,
  query = "SELECT ft_seg_stress, tf_seg_stress, geom FROM received.sa_ways"
)

bna_pal <- c("#FC7151","#DC7E6A","#C98875","#C08B83","#AD9396",
             "#9C9A9F","#929EAC","#78AAC5","#6FADCB","#49BFE6")
bna_breaks <- c(6,12,18,24,30,36,42,48,54,100)
ways$ft_stress <- ifelse(ways$ft_seg_stress == 1,"low stress","high stress")
ways$tf_stress <- ifelse(ways$tf_seg_stress == 1,"low stress","high stress")
library(tmap)
tmap_mode("view")

tmap mode set to interactive viewing

int_map <- 
  tmap::tmap_leaflet(
     tmap::tm_view(
     basemaps = c(
       "CartoDB.Positron",
       "CartoDB.DarkMatter",
       "OpenStreetMap.Mapnik"
     )
   ) +
     tmap::tm_shape(pop) +
     tmap::tm_polygons(
       col = "overall_score",
       style = "fixed",
       breaks = bna_breaks,
       palette = bna_pal,
       alpha = 0.8,
       title = "BNA score",
       border.col = NULL,
       colorNA = NULL,
       showNA = FALSE
      ) +
     tmap::tm_shape(ways) +
     tmap::tm_lines(
       col = "ft_stress", 
       colorNA = NULL,
       showNA = FALSE,
       palette = c("firebrick1", "deepskyblue3"),
       title.col = "Stress network"
      ) +
     tmap::tm_shape(ways) +
     tmap::tm_lines(
       col = "tf_stress", 
       colorNA = NULL,
       showNA = FALSE,
       palette = c("firebrick1", "deepskyblue3"),
       legend.col.show = FALSE
      )
  )
int_map

Observations

• The total time that this particular city took to compute its BNA, including plots and data download was:

Time difference of 30.40046 mins

• What can be observd for the whole analysis is that the resulting BNA score is highly influenced by the fact that the job/employment data is not available. However, this was an attempt to reproduce the score as close as possible as PfB apply their methodology, just to explore its reproducibility.

• My plan next is to exclude this variable from the BNA score computation, and perhaps include some other variables that would suit the European context better.

• My final goal for the moment is to try to validate the scoring methodology for Europe. I picked a city in the UK as I know there is Origin-Destination data available that could be used as a validation method.